Actuation systems exist in nature, but the most complex mechanisms you see in the wild are often difficult to re-create. Some natural actuators perform multiple functions at the same time.

Take the cuttlefish, for example. The shape-shifting sea creatures will contract their muscles and nodules (known as papillae) to camouflage themselves, modulating skin color and surface simultaneously.

“Part of the reason why we are unable to replicate this in artificial devices is that we find it extremely challenging to create and optimize such designs,” said Subramanian Sundaram, a former graduate student in MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). “And even when we do so, it remains challenging to manufacture them.”

Sundaram and his team set out to see whether it was possible to create sophisticated, multi-functional actuators by generating designs automatically.

And what’s more complex than recreating art on an actuator?

In a paper published this week  in Science Advances, the researchers fabricated an actuator  that showed a Vincent van Gogh portrait when laid flat, and displayed the famous Edvard Munch painting “The Scream” when activated and tilted at an angle.

an actuator from MIT shows a Van Gogh portrait when flat and The Scream from Edvard Munch when tilted

In addition to art, the researchers also 3D-printed floating water lilies. The flower petals, equipped with arrays of actuators and hinges, fold up in response to magnetic fields run through conductive fluids.

The actuators are made from a patchwork of three components: a near-transparent rigid material, an opaque flexible material used as a hinge, and a brown nanoparticle material that responds to a magnetic signal.

To create the shape-shifting object, the range of possible characterization data – color, magnetization, and rigidity – was placed into a property library for simulation.

Sundaram's software first breaks down the actuator design into millions of three-dimensional pixels, or “voxels,” which can each be filled with any of the materials.

Then, the software runs millions of simulations, filling voxels with different materials and eventually landing on the optimal placement to generate two different images at two different angles: Van Gogh and Edvard Munch, for example.

It’s a bit like a Rubik’s Cube, except 5.5 million voxels are being iteratively reconfigured to match an image and meet a measured angle.

“We’re comparing what that [voxel column] looks like when it’s flat or when it’s tilted, to match the target images,” said Sundaram. “If not, you can swap, say, a clear voxel with a brown one. If that’s an improvement, we keep this new suggestion and make other changes over and over again.”

To compute the actuator’s appearances at each iteration, the researchers adopted a computer graphics technique called “ray-tracing,” which simulates the path of light interacting with objects. Simulated light beams shoot through the actuator at each column of voxels.

A custom 3D printer then fabricates the actuator by dropping 30-micron-sized droplets of the determined material into the right voxel, layer by layer.

3D-printed water lillies from MIT. Arrays of actuators and hinges fold up in response to magnetic fields run through conductive fluids
The 3D-printed floating water lilies. The petals are equipped with arrays of actuators and hinges that fold up in response to magnetic fields run through conductive fluids. (Image Credit: Subramanian Sundaram)

The MIT team's work could be used someday to support the design of larger structures like airplane wings, according to the report's lead author.

“Our ultimate goal is to automatically find an optimal design for any problem, and then use the output of our optimized design to fabricate it,” said Sundaram.

Sundaram spoke with Tech Briefs about what is possible when you can design an actuator with 5.5 million variables.

Tech Briefs: What is challenging about actuator design that your system is meant to address?

Subramanian Sundaram: The challenging aspect of any actuator design is the number of choices a designer has to make. Let's consider a simple example: If you were asked to design a simple brick, you have to decide the length, width, and height. In addition, you have to choose the material that the brick is made of. In this simple case, these are all the variables in the design, or the dimensions of this problem.

When it comes to designing real world actuators – say, a jet engine – then the number of these design dimensions is extremely large. Typically, the materials and geometry of each part of the system are carefully tuned by experienced specialists to achieve the desired performance. This process of coming up with a design is painstaking and usually slow. In that sense, a system that can generate complex designs and fabricate them would be valuable.

The actuator design we chose to demonstrate has 5.5 million dimensions. Essentially, the optimizer has to choose a single material for each cell in a grid of 5.5 million cells. In our case, we tell our algorithm that we require an actuator that looks like the image of Van Gogh, but when a magnetic field is applied, it tilts and looks like "The Scream." Once the optimizer generates the final design, our 3D-printer fabricates the actual structure with the right materials in each cell.

Tech Briefs: Why is your design method better than designing actuators by hand?

Sundaram: When you have to make a large number of design choices, it is nearly impossible for a human to design such parts. Furthermore, it is unlikely your design is good, or optimal. In contrast, our optimizer comes up with a design, rapidly tests if the design is good and modifies it quickly to a newer design and tests it repeatedly. Sometimes, reaching the optimized design could take millions of tries, so it is often not possible to do this by hand.

Tech Briefs: What criteria does the system use to make its automated decisions?

Sundaram: Our system is required to generate a design that looks like the image of Van Gogh. When actuated by a magnetic field, however, it is required to tilt and look like a different image. So, the main criteria are the appearance (images) and deflection.

The system's main task is to produce the right appearance and the correct deflection when a magnetic field is applied. First, an actuator design is randomly generated, and its appearance is compared to desired images. Naturally, the first guess is pretty bad. The system then mildly tweaks the design and the evaluates the performance. If this newer design is better, it is retained. If not, it is discarded. After millions of such small tweaks, the final actuator design is actually pretty good.

Tech Briefs: What kinds of applications is this system ideal for?

Sundaram: This system is ideal for designs that are hard to be optimized, and where the number of design choices is large.

Tech Briefs: What is most exciting to you about this system and what is possible?

Sundaram: The general concept of computational morphogenesis (i.e., evolving an optimal design automatically) is quite powerful.

When we design parts manually, we are more inclined to work with regular shapes, but the optimal design need not be a simple or regular structure. So, computationally generating a design is less biased in that sense.

In parallel, emerging additive manufacturing processes offer unprecedented freedom in fabrication and increasingly in material choices. Combining these two ideas is what I'm most excited about, and our study is fundamentally aimed at linking these two aspects together.

When more functional materials are available, then this concept becomes even more powerful. This is by no means an easy task, but miniaturized actuators are pervasive across a wide range of applications today ranging from robotics to space technologies to medicine. So, I see this as a foundational and compelling problem to work on.

What do you think? Share your comments and questions below.